LIVE
CliffCompaction: A Context-Compaction Method to Cut Costs of Coding Agents22/09/26|FleXray: A Generalist Model for Whole-Body X-ray Segmentation22/09/26|Anthropic unveils Claude Opus 5.522/09/26 · Anthropic|Court filings reveal internal admissions on AI content scraping19/09/26|US Military Avoids Close Call After AI-Generated Intelligence Hallucination18/09/26|Study finds gender bias in GPT models is not reduced but reshaped across generations17/09/26 · OpenAI|Google DeepMind adds secure server-side memory to Private AI Compute23/09/26 · Google DeepMind|Google DeepMind adds text-to-speech to Gemini 3.823/09/26 · Google DeepMind|NVIDIA releases Nemotron 3 Diarization for real-time speaker identification23/09/26 · NVIDIA|OpenAI extends its Daybreak cyber program to Ukraine for civilian defense23/09/26 · OpenAI|OpenAI enlists influencers to burnish its public image23/09/26 · OpenAI|Claude Code Reportedly Reads AGENTS.md Only When Telemetry Is Enabled23/09/26 · Anthropic|CliffCompaction: A Context-Compaction Method to Cut Costs of Coding Agents22/09/26|FleXray: A Generalist Model for Whole-Body X-ray Segmentation22/09/26|Anthropic unveils Claude Opus 5.522/09/26 · Anthropic|Court filings reveal internal admissions on AI content scraping19/09/26|US Military Avoids Close Call After AI-Generated Intelligence Hallucination18/09/26|Study finds gender bias in GPT models is not reduced but reshaped across generations17/09/26 · OpenAI|Google DeepMind adds secure server-side memory to Private AI Compute23/09/26 · Google DeepMind|Google DeepMind adds text-to-speech to Gemini 3.823/09/26 · Google DeepMind|NVIDIA releases Nemotron 3 Diarization for real-time speaker identification23/09/26 · NVIDIA|OpenAI extends its Daybreak cyber program to Ukraine for civilian defense23/09/26 · OpenAI|OpenAI enlists influencers to burnish its public image23/09/26 · OpenAI|Claude Code Reportedly Reads AGENTS.md Only When Telemetry Is Enabled23/09/26 · Anthropic|
ToolOpenAI

OpenAI improves prompt caching for GPT-6

OpenAI has rolled out improvements to prompt caching for GPT-6, including higher cache hit rates, new diagnostic tools, and explicit breakpoints to cut latency and costs.

September 22, 20263 min readPublished byOpenAI

OpenAI has announced updates to how prompt caching works for GPT-6, a feature aimed at developers who repeatedly send requests containing identical chunks of text, such as system instructions or large context documents. The company says it has raised the cache hit rate, meaning more requests now benefit from faster, cheaper processing instead of being fully reprocessed on every call.

Among the changes are new diagnostic tools that let engineering teams see when and why caching is or isn't being applied. This added visibility should help teams fine-tune applications that rely heavily on the API, particularly those involving long contexts or frequent calls sharing a common prefix, such as conversational agents or retrieval-augmented systems.

OpenAI is also introducing explicit breakpoints, giving developers precise control over where caching should stop within a prompt. This finer-grained control offers more predictability over system behavior, which matters for production applications where latency and cost need to be managed reliably rather than left to chance.

The announcement fits into a broader industry trend of language model providers working to lower the operating cost of applications built around long contexts. Anthropic and Google have offered comparable caching mechanisms for some time, and this update brings GPT-6 closer to parity with competitors on this specific front. While the improvement is primarily a technical one aimed at engineering teams, it can directly affect the margins of companies running AI products at scale.

Tags
prompt-cachinggpt-6apilatencycost-optimizationdeveloper-tools

Read also

OpenAI improves prompt caching for GPT-6 · nAIvigate